Moreno-Romero et al. Genome Biology (2019) 20:41 https://doi.org/10.1186/s13059-019-1652-0 RESEARCH Open Access Epigenetic signatures associated with imprinted paternally expressed genes in the Arabidopsis endosperm Jordi Moreno-Romero1,2, Gerardo Del Toro-De León2, Vikash Kumar Yadav2, Juan Santos-González2 and Claudia Köhler2* Abstract Background: Imprinted genes are epigenetically modified during gametogenesis and maintain the established epigenetic signatures after fertilization, causing parental-specific gene expression. Results: In this study, we show that imprinted paternally expressed genes (PEGs) in the Arabidopsis endosperm are marked by an epigenetic signature of Polycomb Repressive Complex2 (PRC2)-mediated H3K27me3 together with heterochromatic H3K9me2 and CHG methylation, which specifically mark the silenced maternal alleles of PEGs. The co-occurrence of H3K27me3 and H3K9me2 on defined loci in the endosperm drastically differs from the strict separation of both pathways in vegetative tissues, revealing tissue-specific employment of repressive epigenetic pathways in plants. Based on the presence of this epigenetic signature on maternal alleles, we are able to predict known PEGs at high accuracy and identify several new PEGs that we confirm using INTACT-based transcriptomes generated in this study. Conclusions: The presence of the three repressive epigenetic marks, H3K27me3, H3K9me2, and CHG methylation on the maternal alleles in the endosperm serves as a specific epigenetic signature that allows prediction of genes with parental-specific gene expression. Our study reveals that there are substantially more PEGs than previously identified, indicating that paternal-specific gene expression is of higher functional relevance than currently estimated. The combined activity of PRC2-mediated H3K27me3 together with the heterochromatic H3K9me3 has also been reported to silence the maternal Xist locus in mammalian preimplantation embryos, suggesting convergent employment of both pathways during the evolution of genomic imprinting. Keywords: H3K9me2, H3K27me3, CHG methylation, Endosperm, Genomic imprinting, Paternally expressed genes Background central cell to give rise to the triploid endosperm [2]. Genomic imprinting is an epigenetic phenomenon causing Imprinted genes are epigenetically modified during gam- maternal and paternal alleles to be differentially expressed ete formation, and the established epigenetic asymmetry is after fertilization. In plants, genomic imprinting is mainly maintained after fertilization. Differential DNA methyla- confined to the endosperm, an ephemeral nutritive tissue tion is established by the DNA glycosylase DEMETER supporting embryo growth, similar to the placenta in (DME) that removes methylated cytosine residues in the mammals [1]. The endosperm is the product of a double central cell of the female gametophyte [3]. DME is not ac- fertilization event, where one of the haploid sperm cells tive in sperm, leading to differential DNA methylation be- fertilizes the haploid egg cell giving rise to the diploid em- tween male and female genomes in the endosperm. DME bryo, while the second sperm cell fertilizes the diploid acts on small transposable elements (TEs) in the vicinity of genes [4] and its activity has been connected to the ex- pression of maternally expressed imprinted genes (MEGs). * Correspondence: claudia.kohler@slu.se 2 Department of Plant Biology, Uppsala BioCenter, Swedish University of Hypomethylation can furthermore cause repression [5, 6], Agricultural Sciences and Linnean Center for Plant Biology, Uppsala, Sweden possibly by exposing binding sites for the Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Moreno-Romero et al. Genome Biology (2019) 20:41 Page 2 of 11 Fertilization-Independent Seed (FIS)-Polycomb Repressive that the presence of both modifications on the maternal Complex 2 (PRC2) [7, 8], an evolutionary conserved chro- alleles correlates with paternally biased expression. Con- matin modifying complex that applies a trimethylation sistent with this notion, we found that genes previously mark on histone H3 at lysine 27 (H3K27me3) [9]. The identified as PEGs [14] were substantially enriched for Arabidopsis FIS-PRC2 consists of the subunits MEDEA both modifications on their maternal alleles, while (MEA), FIS2, FERTILIZATION-INDEPENDENT ENDO- MEGs did not show enrichment of both marks on either SPERM (FIE), and MULTICOPY SUPPRESSOR OF IRA1 maternal or paternal alleles (Fig. 1c, d, Additional file 1: (MSI1) [10] and is specifically active in the central cell of Figure S1C-D). This trend was independent of the direc- the female gametophyte and in the endosperm [11]. Re- tion of the cross and similarly observed in Col × Ler and pression of the maternal alleles of PEGs is mediated by Ler × Col crosses (Fig. 1 and Additional file 1: Figure the activity of the FIS-PRC2 [12], consistent with maternal S1). The presence of H3K27me3 is generally confined to PEG alleles being marked by H3K27me3 [8]. In this gene bodies [15], consistent with the observed enrich- manuscript, we addressed the mechanism of maternal al- ment of H3K27me3 in gene bodies of PEG maternal al- lele repression in PEGs. We surprisingly found that PEGs leles (Fig. 1e and Additional file 1: Figure S1E). are regulated by two otherwise largely exclusive epigenetic Interestingly, H3K9me2 was similarly restricted to gene repressive pathways, the FIS2-PRC2 and the pathway es- bodies of PEGs (Fig. 1e and Additional file 1: Figure tablishing the heterochromatin-localized H3K9me2 modi- S1E), contrasting the exclusion of H3K9me2 from genic fication [13]. We demonstrate that both modifications, regions in sporophytic tissues [16]. The maternal alleles H3K27me3 and H3K9me2, overlap on the maternal alleles of PEGs were also strongly enriched for CHG methyla- of the majority of PEGs. Our data suggest that most likely tion (Fig. 1f, Additional file 1: Figure S1 and S2), consist- FIS-PRC2 acts first and is required to establish H3K9me2. ent with the chromomethylase 3 (CMT3) acting in a Furthermore, we find maternal alleles of PEGs to be positive feedback loop with the histone methyltransfer- marked by CHG methylation in the central cell, indicating ase proteins KYP/SUVH4, SUVH5, and SUVH6 [13]. that repressive pathways establishing H3K27me3, Levels of CHH methylation were generally higher on H3K9me2, and CHG methylation act in the central cell of maternal than on paternal alleles (Fig. 1f, Additional file 1: the female gametophyte. Finally, we use the presence of Figure S1F), consistent with depletion of CHH methyla- the three modifications to predict novel PEGs and propose tion in sperm [4, 17]. that the number of PEGs predicted based on expression data strongly underestimates the real number of PEGs. Maternal-specific CHG methylation is established in the central cell and depends on FIS-PRC2 Results PEGs had increased levels of CHG methylation in the The maternal alleles of PEGs are marked by H3K27me3, central cell, but low CHG methylation in sperm and H3K9me2, and CHG methylation vegetative cells of pollen (Fig. 2a), revealing that differ- In a previous study, we revealed that H3K27me3 and ences in CHG methylation are established before H3K9me2 overlapped in the endosperm at pericentro- fertilization. The FIS-PRC2 had been proposed to pro- meric heterochromatic regions of the paternal genome, mote non-CG methylation [4], raising the hypothesis suggesting a partial functional redundancy of both modi- that increased CHG methylation on the maternal alleles fications [8]. We now addressed the question whether of PEGs depends on the activity of FIS-PRC2 in the cen- this redundancy extends to other regions of the genome. tral cell. Consistently, we observed strongly decreased We found that about one third of genes marked by CHG methylation on the maternal alleles of PEGs in H3K27me3 on the maternal alleles were also marked by seeds lacking maternal FIE activity (Fig. 2b), suggesting H3K9me2 (Fig. 1a, hypergeometric test, P = 0 (Col × Ler that FIS-PRC2 activity is required to recruit CHG crosses); Additional file 1: Figure S1A, P = 0 (Ler × Col methylation. The DNA glycosylase DME is required for crosses)). Genes containing both modifications on their the activation of MEA and FIS2, both encoding subunits maternal alleles had significantly higher levels of of the FIS-PRC2 [18, 19]. Maternal PEG alleles had re- H3K27me3 compared to those only marked by duced levels of CHG methylation in seeds lacking mater- H3K27me3 (Fig. 1b, Additional file 1: Figure S1B). The nal DME activity (Fig. 2b), supporting the notion that majority of double-marked genes contained both modifi- FIS-PRC2 activity is required for CHG methylation es- cations specifically on the maternal but not the paternal tablishment on maternal PEG alleles. alleles (maternal-specific marks) (Fig. 1a, Additional file 1: Previous work revealed that loss of FIS-PRC2 function Figure S1A) and increasing levels of H3K27me3 on ma- caused activation of maternal PEG alleles [12]. We ad- ternal alleles correlated with increasing levels of dressed the question whether loss of CMT3 and the re- H3K9me2 on the maternal but not the paternal alleles dundantly acting SUVH4,5,6 [20] similarly affect (Additional file 1: Figure S2 and S3). We thus proposed silencing of maternal PEG alleles. Based on published Moreno-Romero et al. Genome Biology (2019) 20:41 Page 3 of 11 A B D C E F Fig. 1 H3K27me3, H3K9me2, and CHG methylation overlap on maternal alleles of PEGs in crosses Col × Ler. a Overlap of genes marked by H3K27me3 and H3K9me2 on the maternal alleles. Dark red inset marks genes that have H3K27me3 and H3K9me2 exclusively on their maternal alleles (excluding genes also marked on the paternal alleles). b Genes marked by H3K27me3 and H3K9me2 have higher levels of H3K27me3. Box plots show mean values of z-scored H3K27me3 and H3K9me2 of genes where the maternal alleles are either marked by H3K27me3 alone or by a combination of H3K27me3 and H3K9me2. Genes containing maternal-specific H3K27me3 and H3K9me2 are particularly enriched for H3K27me3. Boxes show medians and the interquartile range, error bars show the full range excluding outliers. Notches show the 95% confidence interval for the median. Asterisks mark significance (Wilcoxon rank sum test, P < 1.0e−3). Presence of the mark corresponds to the mark highest score defined in Additional file 2: Table S1. c The majority of PEGs are marked by H3K27me3 and H3K9me2 on their maternal alleles. Overlap of PEGs [14] with maternal and paternal H3K27me3 and H3K9me2. d Levels of H3K27me3 and H3K9me2 on PEGs, MEGs, and all genes. Box plots (as specified in panel B) show mean values of z-scored H3K27me3 and H3K9me2. Asterisks mark significance (Wilcoxon rank sum test, P < 1.0e−3). e Metagene plots showing average distribution of z-score normalized H3K27me3 and H3K9me2 levels on maternal and paternal alleles of PEGs [14]. f Levels of DNA methylation in each sequence context (mCG, mCHG, and mCHH) on PEGs, MEGs, and all genes. Box plots (as specified in panel B) show mean values of relative DNA methylation. Asterisks mark significance (Wilcoxon rank sum test, P < 5.0e−3). Data shown correspond to cross Col × Ler [8], for reciprocal cross direction see Additional file 1: Figure S1 Moreno-Romero et al. Genome Biology (2019) 20:41 Page 4 of 11 A B Fig. 2 Maternal CHG methylation is established in the central cell and depends on FIS-PRC2. a Box plots showing mean values of relative DNA methylation in the central cell, sperm cell, and vegetative cell for PEGs, MEGs, and all genes. Boxes show medians and the interquartile range, error bars show the full range excluding outliers. Notches show the 95% confidence interval for the median. Asterisk marks significance (Wilcoxon rank sum test, P < 5.0e−3). b Box plots showing mean values of CHG methylation for PEGs and MEGs on maternal and paternal alleles in wild- type, fie (left panels) and dme (right panels) endosperm. Asterisk marks significance (Wilcoxon rank sum test, P < 5.0e−2) expression studies, CMT3, SUVH4, and SUVH6 are alleles did not shift the allelic balance, the combination expressed in the central cell of the female gametophyte of more than one modification shifted the balance to- and in the endosperm [21, 22]. The maternal alleles of wards preferential paternal allele expression (Fig. 3a, seven tested PEGs remained silenced in reciprocal Additional file 1: Figure S5A), suggesting that the com- crosses of wild type with cmt3 and suvh4,5,6 triple mu- bined presence of H3K27me3, H3K9me2, and CHG tants (Additional file 1: Figure S4A-B), indicating that methylation is a hallmark for maternal allele repression. stable silencing of the maternal alleles of PEGs does not Consistently, increased bias towards paternal allele ex- depend on CMT3 and SUVH4,5,6 activity before pression correlated with increased levels of H3K27me3, fertilization. We tested the requirement of CMT3 and H3K9me2, and CHG methylation on maternal alleles SUVH4,5,6 for PEG regulation after fertilization by (Fig. 3b, Additional file 1: Figure S5B). monitoring PEG expression in homozygous mutant We tested which combination of marks would most reli- cmt3 and suvh4,5,6 seeds. None of the seven tested ably allow predicting paternally biased genes. We assigned genes was significantly upregulated in seeds of cmt3 or scores for the allele-specific presence of the three modifi- suvh4,5,6 mutants (Additional file 1: Figure S4C), indi- cations (see Additional file 2: Table S1) and tested whether cating that CMT3 and SUVH4,5,6 are not required for highly scoring genes were enriched for previously pre- repression of maternal PEG alleles after fertilization. In dicted PEGs [14]. The combination of CHG methylation the suvh4,5,6 triple mutant, H3K9me2 is genome-wide in the central cell together with maternal-specific eliminated in vegetative tissues [23]; however, whether H3K27me3 and H3K9me2 allowed to predict the highest H3K9me2 is similarly eliminated in the central cell and number of previously described PEGs in Col and Ler ac- endosperm remains to be shown. Therefore, deciphering cessions (24 out of 42 (57%), [14]) in relation to the num- the role of H3K9me2 in repressing the maternal alleles ber of genes in the category with the highest score of PEGs remains to be subject of future investigations. (Fig. 4a, Additional file 2: Table S1) and was chosen for further analysis. The category with the highest score was Paternally biased expression coincides with the significantly enriched for PEGs (hypergeometric test, P = combination of H3K27me3, H3K9me2, and CHG 1.0e−32), while categories with lower scores contained methylation only few PEGs (Fig. 4a). Similarly, out of 64 PEGs that had We addressed the question whether the presence of been predicted by a recent study re-evaluating previously H3K9me2 and CHG methylation is functionally relevant published imprintome datasets [24], 40 (62%) PEGs were for maternal allele repression by testing whether the present in the highest score category (Fig. 4a). Nearly half presence of both modifications either alone or in com- (96 genes, 46.4%) of those genes in the highest score cat- bination with H3K27me3 correlates with maternal allele egory were significantly paternally biased (chi-square < repression. Strikingly, while the presence of either CHG 0.05, Bonferroni corrected, Additional file 3: Table S2), methylation, H3K9me2, or H3K27me3 on maternal which was significantly more than the 8% paternally Moreno-Romero et al. Genome Biology (2019) 20:41 Page 5 of 11 A B Fig. 3 Paternally biased expression coincides with the combination of H3K27me3, H3K9me2, and CHG methylation in crosses Col × Ler (for reciprocal cross direction see Additional file 1: Figure S5). a Box plots show mean values of maternal to total reads for genes marked by one or several of the following modifications: CHG methylation in the central cell, difference of CHG methylation, H3K27me3, or H3K9me2 levels on maternal and paternal alleles in the endosperm. For each modification only genes present in the highest scoring category were included in the analysis (see Additional file 2: Table S1). Boxes show medians and the interquartile range; error bars show the full range excluding outliers. Widths of boxes are proportional to square-root of the number of genes in each category. Blue line indicates the expected ratio for biallelically expressed genes. b Paternal expression bias correlates with high levels of CHG methylation in the central cell and H3K27me3 and H3K9me2 on maternal alleles in the endosperm. Plotted are mean values of CHG methylation, H3K27me3, and H3K9me2 on maternal and paternal alleles for genes with deviating parental expression levels from the expected two maternal to one paternal ratio biased genes identified among all genes tested (hypergeo- have been classified as biallelically expressed or mater- metric test, P = 8.9e−52). We thus conclude that the pres- nally biased are more highly expressed in the seed coat ence of the three modifications, CHG in the central cell than those with paternally biased expression. We tested and H3K27me3 and H3K9me2 on maternal alleles in the this hypothesis using available expression data of seed endosperm, allows to predict genes with paternally biased coat tissue [22] at a similar developmental stage as pub- expression. Paternally biased genes were particularly in- lished endosperm expression data [14]. Indeed, genes volved in regulation of transcription (P = 7.54e−5) and with paternally biased expression were significantly less chromatin modification (P = 1.57e−3) (Additional file 1: expressed in the seed coat compared to genes that based Table S3), consistent with previous reports on the func- on available expression data are either predicted to be tional role of PEGs [14, 25]. biallelically expressed or maternally biased (Wilcoxon rank sum test, P < 0.005; Fig. 4b), strongly suggesting Maternal seed coat contamination restricts the that maternal tissue contamination in available expres- identification of PEGs sion data limits the prediction of PEGs. To further test Published endosperm transcriptome data contain a sub- this hypothesis, we generated transgenic reporter lines stantial fraction of transcripts from the maternal seed containing the promoter (≃ 2 kb) and downstream genic coat, which may limit the correct prediction of pater- regions fused to the green fluorescent protein (GFP) of nally biased genes [24]. We hypothesized that there are seven genes belonging to the highest score category but several genes that based on their epigenetic modifica- predicted to be maternally (AT2G33620, AT1G43580, tions (score 12, Fig. 4a and Additional file 3: Table S2) AT1G47530, AT1G64660, AT2G30590, AT4G15390) or are likely to be PEGs but based on available expression biallelically expressed (AT5G53160). We detected a GFP data are not correctly classified. A prediction of this hy- signal in the endosperm only for construct AT1G64660 pothesis is that genes in the highest score category that (Fig. 5, Additional file 1: Table S4). For construct Moreno-Romero et al. Genome Biology (2019) 20:41 Page 6 of 11 A B Fig. 4 Epigenetic signatures allow predicting paternally expressed imprinted genes (PEGs). a The combination of CHG methylation in the central cell together with maternal-specific H3K27me3 and H3K9me2 allows predicting previously described PEGs [14, 24]. Scores are calculated as described in the Additional file 2: Table S1. High and low scores correspond to high or low levels of CHG methylation, H3K27me3, and H3K9me2, respectively. Number of PEGs corresponds to PEGs previously predicted [14] (i) and [24] (ii) in Col and Ler accessions. Total genes correspond to all genes present in the corresponding categories. Box plots show mean values of maternal to total reads in the indicated categories. Boxes show medians and the interquartile range, error bars show the full range excluding outliers. Blue line indicates the expected ratio for biallelically expressed genes. b Box plots show relative expression level of genes in the seed coat of torpedo stage embryos [22]. Genes present in the highest score category (Additional file 3: Table S2) were grouped in paternally biased genes (chi-square test, P < 0.05 Bonferroni corrected) and not-paternally biased genes. Asterisks mark significance (Wilcoxon rank sum test, P < 5.0e−3) AT1G47530, a signal was detected in the seed coat, while support the hypothesis that seed coat contamination for the other constructs no GFP signal was detected in limits the transcriptome-based identification of PEGs. seeds, indicating that the regulatory elements required To test whether we could confirm additional PEGs for the expression of those genes are located outside the that have been predicted based on their epigenetic signa- promoter and genic regions used to generate the re- tures, we applied INTACT (isolation of nuclei tagged in porter lines. Reciprocal crosses using the AT1G64660 re- specific cell types) to purify endosperm nuclei at 4 days porter lines revealed that this gene is indeed a PEG and after pollination (DAP) from Col × Ler and Ler × Col re- strongly expressed in the endosperm when paternally, ciprocal crosses. Isolated RNA was sequenced and pro- but not when maternally inherited (Fig. 5). These data filed for allele-specific gene expression. By analyzing the Fig. 5 Reporter lines for AT1G64660 confirm imprinted expression. Representative images of seeds derived after reciprocal crosses of the AT1G64660 reporter line (fusion with green fluorescent protein (GFP)) with wild-type (WT) plants. GFP fluorescence was detected in the seed coat when the AT1G64660 reporter was maternally inherited, but endosperm-specific expression was only detected when the AT1G64660 reporter was paternally inherited. Seeds at 2 DAP were used for imaging. Red stain is propidium iodide. Scale bars correspond to 50 μm Moreno-Romero et al. Genome Biology (2019) 20:41 Page 7 of 11 maternal to total reads ratio in each epigenetic category, of FIS-PRC2 function [12] suggests that H3K9me2 and we confirmed that the genes in the highest score cat- CHG methylation require FIS-PRC2 function. The PRC2 egory (group with score 12) indeed showed a clear trend is generally targeting genes with specific roles during de- towards paternally biased expression (Fig. 6a, Add- velopment [30], while complexes establishing H3K9me2 itional file 4: Table S5). Following previously established are mainly targeting TEs localized in heterochromatic criteria [14], we predicted 148 PEGs that were recipro- regions of the genome [31]. This functional division of cally imprinted in both directions of the crosses. There PRC2 and machineries establishing heterochromatic was a significantly higher number of PEGs present in marks is conserved in plants as well as in mammals [32]; the highest score category compared to a representative however, there are notable exceptions to this rule in random sample of genes with informative reads (Fig. 6b). both groups of organisms. In rice seedlings, about one Furthermore, the highest score category had the highest third of H3K27me3 marked genes are also marked by frequency of PEGs, with other categories having signifi- CHG and CHH methylation [33]. Similar to the findings cantly fewer PEGs (Fig. 6c). Of the 148 genes that we reported in our study, higher levels of H3K27me3 correl- predicted as PEGs based on our RNA sequencing data, ate with higher CHG and CHH methylation in gene 45 were present in the highest score category, which is bodies of rice [33]. The rice H3K27me3 methyltransfer- significantly more than expected by chance (P = 2.008e ase SDG711 physically interacts with the CHH methyl- −55, Fig. 6d). Out of those, 24 were previously predicted transferase OsDMR2 and the SRA-domain containing based on published data [12, 14, 26, 27], while 21 genes SUVH protein SDG703, uncovering a mechanistic con- are likely new high-confidence PEGs, revealing that nection between PRC2 and non-CG methylation. PEGs are more common than previously estimated. Which methyltransferases establish H3K9me2 in the central cell of the female gametophyte and in the endo- Discussion sperm of Arabidopsis remains to be investigated. In this study, we identified the concomitant presence of SUVH4,5,6 are the main H3K9me2 methyltransferases maternal-specific CHG methylation, H3K27me3, and in sporophytic tissues of Arabidopsis [23] and SUVH4 H3K9me2 as an epigenetic signature for paternally and SUVH6 are expressed in the central cell and in the biased expression in the endosperm. We furthermore endosperm [21, 22]. However, the imprinted genes predict that there are substantially more PEGs than pre- SUVH7 and SUVH8 [12] encode for two potential viously reported in Arabidopsis [12, 14, 26], suggesting H3K9me2 methyltransferases that despite lack of in vitro that in Arabidopsis the number of PEGs exceeds the activity [20] may be active in the endosperm. Increased number of MEGs. Recent re-evaluation of published expression of SUVH7 is detrimental in triploid seeds imprintome data of Arabidopsis revealed that a large [34], indicating that SUVH7 is functionally active in the number of previously predicted MEGs were seed coat endosperm. Similarly, CMT3 is expressed in the central expressed genes, while the number of PEGs was under- cell and in the endosperm [21, 22]; however, functional estimated [24]. Our study supports and extends this no- redundancies with other CMT genes cannot be ruled out tion by showing that there is a large number of based on available data. Identifying the H3K9me2 and paternally biased genes that likely failed to be identified CHG methyltransferases acting in the endosperm will be in previous studies because of maternal seed coat con- a major step to address the functional role of H3K9me2 tamination or early stage-specific expression. and CHG methylation in the stable repression of mater- A previous study reported that the maternal alleles of nal PEG alleles. PEGs in A. lyrata are marked by CHG methylation and In mammalian cells, H3K9 methyltransferases colocal- implicated that closely related species use different mecha- ize with PRC2 [35–37], revealing crosstalk between these nisms to regulate imprinted gene expression [28]. Our two major epigenetic silencing pathways that likely is re- study reveals that similar to A. lyrata the maternal alleles quired for stable gene silencing. The imprinted maternal of many PEGs are also marked by CHG methylation in A. Xist locus encoding an X-linked long-noncoding RNA is thaliana, highlighting that epigenetic mechanisms covered by H3K27me3 and H3K9me3 in preimplanta- employed to maintain monoallelic expression are rather tion embryos [38, 39]. Importantly, loss of maternal conserved between related species. Interestingly, while the H3K27me3 induces Xist activation, indicating that ma- maternal alleles of PEGs in maize are marked by ternal H3K27me3 is the major imprinting mark of Xist H3K27me3, they are not marked by CHG methylation [39]. This is strikingly similar to findings made in this [29], indicating that diverged species may use a different study revealing H3K27me3 as the major repressive mark mechanism in maintaining maternal allele repression. for PEGs. Recent work revealed that maternal How H3K9me2 and CHG methylation are established H3K27me3 controls DNA methylation-independent im- at PRC2 target genes remains to be studied; however, printing in mammalian preimplantation embryos [40]. the strong activation of maternal PEG alleles upon loss While imprinted expression of most genes is lost in the Moreno-Romero et al. Genome Biology (2019) 20:41 Page 8 of 11 A B C D Fig. 6 Confirmation of predicted PEGs in RNA samples of INTACT-purified endosperm nuclei. a Distribution of maternal to total read values for genes in each epigenetic category. The distribution for the total population of genes with informative reads (U) is included. Genes with the highest levels of CHG methylation, H3K27me3, and H3K9me2 show a clear trend towards paternally biased expression (around 0.5). Values correspond to combined values of Col × Ler and Ler × Col crosses. Boxes show medians and the interquartile range, lines show the full range of outliers. Blue line indicates the expected ratio for biallelically expressed genes (0.66). b Parent-of-origin distribution of the group of genes with score 12, compared with a random group of the same size. Significantly more PEGs are detected in the group of genes with score 12, compared with the random sample (hypergeometric test, P values correspond to the comparison with the total population of PEGs in the whole RNAseq data). The scatter plots show maternal to total read values for Col × Ler and Ler × Col reciprocal crosses. Colors indicate genes with significant parent-of-origin biased expressed: MEGs, 85% or more of maternal reads in each reciprocal cross; PEGs, 50% or more of paternal reads in each reciprocal cross. Maternally and paternally biased genes show significantly biased expression (P ≤ 0.01) in both reciprocal crosses. c Percentage of PEGs, MEGs, and biallelically expressed genes in each epigenetic category. Color code indicates the percentage of genes within each category. Categories with genes having lower levels of CHG methylation, H3K27me3, and H3K9me2 marks have significantly fewer PEGs than the highest scoring category (chi-square test, P < 1e−8 in categories 1–10; P = 0.001 in category 11). Only PEGs, MEGs, and biallelic categories were considered. d Overlap of PEGs that were detected in our INTACT-based RNAseq samples (blue) with genes predicted as PEGs based on their epigenetic score 12 (green) and previously reported PEGs (yellow) by [12, 14, 26, 27]. New predicted and confirmed PEGs in bold. Statistical significance was calculated using a hypergeometric test (P values < 4e−20) embryonic cell lineage, few genes maintain their imprinted expression from those that become bialleli- imprinted expression in the extra-embryonic cell lineage cally expressed remains to be tested, but we consider [40]. The Xist locus that is marked by H3K27me3 and this a very attractive hypothesis. We speculate that the H3K9me3 is among those loci that remain imprinted in presence of both marks in certain tissue types of mam- extra-embryonic tissues [39]. Whether the presence of mals and flowering plants is a conserved epigenetic sig- both marks distinguishes those genes that maintain their nature marking stably repressed genes. Moreno-Romero et al. Genome Biology (2019) 20:41 Page 9 of 11 Conclusions levels of methylation signals in 100-bp bins in the flanks We discovered the co-occurrence of the PRC2-mediated of the genes and in 40 equally long bins between the H3K27me3 and heterochromatic H3K9me2 and CHG transcriptional start and stop. Gene z-scores were calcu- methylation as an epigenetic signature marking the si- lated as an average of z-scores over the gene body. DNA lenced maternal alleles of PEGs. This signature can be methylation data of fie and dme and their corresponding used to predict PEGs at high accuracy, and based on this wild types are from [4]. DNA methylation data of the prediction, we estimate that the number of PEGs is sub- central cell, sperm, and vegetative cells are from [3, 17]. stantially larger than previously estimated. We Endosperm expression data are from [14] and seed coat hypothesize that the common use of PRC2 and H3K9 expression data from [22]. methylation to silence target loci during reproduction has convergently evolved in flowering plants and mam- Endosperm nuclei isolation and RNA sequencing mals to ensure stable silencing during this sensitive life We performed Col × Ler reciprocal crosses using Arabi- stage. dopsis lines expressing PHE1::NTF and PHE1::BirA (lines referred as INT hereafter) [43]. To facilitate the crosses, Methods we used the male sterile mutants pistillata (pi-1, in Ler Plant material and growth conditions accession) and dde2 (in Col accession containing INT) All seeds were surface sterilized (5% sodium hypochlor- as female parents and pollinated them with the INT line ite and 0.01% Triton X-100), stratified for 2 days at 4 °C, (Col accession) and Ler wild type, respectively. A total of and germinated on half-strength Murashige and Skoog 500 mg of siliques for the first replicates and 250 mg of medium containing 1% sucrose under long-day condi- siliques for the second and third replicates were col- tions (16 h light/8 h darkness, 21 °C). Plants were trans- lected at 4 DAP. Tissue homogenization, nuclei purifica- ferred to soil after 10 to 12 days and grown under tion, RNA extraction, and library preparation were long-day conditions. The cmt3-11 (SALK_148381; [41]) performed from three biological replicates as previously and suvh456 mutants (kindly provided by Judith Bender) described [43, 44]. Samples were sequenced at the Na- used in this study are in the Col-0 background. tional Genomic Infrastructure (NGI) from SciLife La- boratory (Uppsala, Sweden) on an Illumina HiSeq2500 Imprinting assays and expression analysis in paired-end 125 bp read length. Mapping and discrim- To generate siliques of indicated crosses, three to five ination of parental reads was done as previously de- flowers were emasculated, hand-pollinated, and harvested scribed [44]. We calculated contamination levels based at 4 DAP. RNA extraction was performed using the Mag- on the deviation of the read counts from the expected JET Plant RNA Purification Kit (Thermo Scientific) follow- 2:1 maternal/paternal genome ratio in the endosperm ing the manufacturer’s instructions. Residual DNA was [43] (Additional file 1: Table S6). Based on this analysis, removed using Invitrogen DNase I (Amplification Grade), the first replicates of both cross directions Col × Ler and and cDNA was synthesized using the Fermentas Ler × Col were not included in the downstream analyses. first-strand cDNA synthesis kit according to the manufac- To increase the statistical power to detect parentally turer’s instructions. Quantitative PCR was performed using biased genes, we merged libraries from two replicates of a MyiQ5 real-time PCR detection system (Bio-Rad) and Col × Ler and three replicates of Ler × Col for down- Solis BioDyne-5x Hot FIREPol EvaGreen qPCR Mix Plus stream analysis. (ROX, Solis BioDyne). For the imprinting-by-sequencing assay, the PCR products were purified and analyzed by Allele-specific expression analysis Sanger sequencing. For the imprinting-by-restriction en- We followed our previously published analysis pipeline zyme digestion assay, the PCR products were purified and to define imprinted genes [44]. Briefly, we defined a digested. Restriction enzymes and primers used are listed in minimum threshold of 20 informative reads for Col × the Additional file 1: Table S6. Ler (2 replicates) and Ler × Col (2 replicates) crosses, re- spectively. Statistical differences between maternal and Data analysis paternal read counts for each gene were calculated using We made use of endosperm-specific ChIP-seq data that a chi-square test, considering genes with a false discov- have been previously generated in our group [8]. Data ery rate adjusted P value of less than 0.01. Additionally, correspond to pooled biological triplicates with ChIP MEGs required to have at least 85% maternal inform- signals being normalized with H3 ChIP data by calculat- ative reads in both directions of the reciprocal cross and ing the log2 ratio in 150-bp bins across the genome. PEGs to have at least 50% paternal informative reads in Data were standardized and normalized with a z-score both directions of the reciprocal cross, following previ- transformation [42]. Metagene plots over genes were ously defined conditions [14]. Quality of sequencing constructed between − 2 and + 2 kb by calculating mean samples is shown in Additional file 1: Table S7. Moreno-Romero et al. Genome Biology (2019) 20:41 Page 10 of 11 Generation of reporter constructs and transgenic lines Funding For the generation of reporter constructs, we used the This research was supported by a European Research Council Starting Independent Researcher grant (to C.K.), a grant from the Swedish Science ClonExpress® MultiS One Step Cloning kit. Promoters Foundation (to C.K.), a grant from the Knut and Alice Wallenberg Foundation (≃ 2 kb) and genic sequences of AT2G33620, (to C.K.), the Göran Gustafsson Foundation for Research in Natural Sciences AT1G43580, AT1G47530, AT1G64660, AT2G30590, and Medicine (to C.K.), and an EMBO fellowship (to G.D.T.D.L). AT4G15390, and AT5G53160 were amplified from WT Availability of data and materials Col-0 genomic DNA using primers specified in Add- The RNA-seq data generated in this study are available through GEO itional file 1: Table S6 and cloned into vector (GSE119915) publicly available at https://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc=GSE119915 [46]. We furthermore used endosperm RNA expres- pB7FWG.0. Constructs were transformed into Agrobac- sion data from [14] (GSE52814), seed coat expression data from [22] terium tumefaciens strain GV3101, and Arabidopsis (GSE12404), parental-specific histone and DNA methylation data from [8] plants were transformed using the floral dip method (GSE66585), central cell DNA methylation profiles from [3] (GSE89789), and DNA methylation data from sperm cell, vegetative cell and endosperm of fie [45]. Ten transgenic lines per construct were generated and dme mutants from [4] (GSE38935). and analyzed. Authors’ contributions JMR, VKY, and GDTDL executed the experimental procedures. JMR, JSG, Imprinting analyses GDTDL, VKY, and CK analyzed the data. JMR and CK wrote the manuscript. All authors discussed the results and commented on the manuscript. All For reciprocal crosses, designated female partners were authors read and approved the final manuscript. emasculated at 1–2 days prior to anthesis. Two days after emasculation, pistils were hand-pollinated with re- Ethics approval and consent to participate Not applicable. spective pollen donors. Seeds were dissected from the si- liques and mounted on a microscope slide for imaging Consent for publication and counting at 2 and 4 DAP. For fluorescence analyses, Not applicable. seeds were stained with 0.1 mg/mL propidium iodide Competing interests (PI) solution in 7% glucose. Seeds of reciprocal crosses The authors declare that they have no competing interests. of reporter lines were analyzed under confocal micros- copy on a Zeiss 800 Inverted Axio Observer with a su- Publisher’s Note persensitive GaASp detector. Images were acquired, Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. analyzed, and exported using Zeiss ZEN software. For each reporter construct, 50–60 seeds per line were Author details 1 analyzed. Present Address: Centre for Research in Agricultural Genomics (CRAG), CSIC-IRTA-UAB-UB, Campus UAB, Bellaterra, Barcelona, Spain. 2Department of Plant Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences Additional files and Linnean Center for Plant Biology, Uppsala, Sweden. Received: 30 October 2018 Accepted: 12 February 2019 Additional file 1: Figure S1. H3K27me3, H3K9me2, and CHG methylation overlap on maternal alleles of PEGs in crosses Ler × Col. Figure S2. Total and parental-specific H3K27me3, H3K9me2, and CHG References methylation levels in the endosperm of previously described PEGs. Fig- 1. Rodrigues JA, Zilberman D. Evolution and function of genomic imprinting ure S3. Increasing levels of H3K27me3 on maternal alleles correlates with in plants. Genes Dev. 2015;29(24):2517–31. increasing levels of H3K9me2. Figure S4. Maternal alleles remain silenced 2. Li J, Berger F. Endosperm: food for humankind and fodder for scientific in suvh4,5,6 and cmt3 mutants. Figure S5. Paternally biased expression discoveries. New Phytol. 2012;195(2):290–305. coincides with the combination of H3K27me3, H3K9me2, and CHG 3. Park K, Kim MY, Vickers M, Park JS, Hyun Y, Okamoto T, Zilberman D, Fischer methylation in crosses Ler × Col (for reciprocal cross direction see Fig. 3). RL, Feng X, Choi Y, et al. DNA demethylation is initiated in the central cells Table S3. GO terms enrichment of genes in the highest score category. of Arabidopsis and rice. Proc Natl Acad Sci U S A. 2016;113(52):15138–43. Table S4. Fluorescence analysis of 4 DAP seeds after reciprocal crosses. 4. Ibarra CA, Feng X, Schoft VK, Hsieh TF, Uzawa R, Rodrigues JA, Zemach A, Table S6. List of primers used in this study. Table S7. Quality of sequen- Chumak N, Machlicova A, Nishimura T, et al. Active DNA demethylation in cing samples. (PDF 1499 kb) plant companion cells reinforces transposon methylation in gametes. Additional file 2: Table S1. Scores assigned to genes based on the Science. 2012;337(6100):1360–4. presence of CHG methylation in the central cell as well as H3K9me2 and 5. Villar C, Erilova A, Makarevich G, Trösch R, Köhler C. Control of PHERES1 H3K27me3 on maternal alleles in the endosperm. (XLSX 1745 KB) (XLSX 1744 kb) imprinting in Arabidopsis by direct tandem repeats. Mol Plant. 2009;2: 654–60. Additional file 3: Table S2. Seed coat expression of genes with 6. Pignatta D, Novitzky K, Satyaki PR, Gehring M. A variably imprinted epiallele maximum score (12) based on Additional file 2: Table S1. (XLSX 45 kb) impacts seed development. PLoS Genet. 2018;14(11):e1007469. Additional file 4: Table S5. Parent-of-origin RNAseq dataset of 4 DAP 7. Weinhofer I, Hehenberger E, Roszak P, Hennig L, Köhler C. H3K27me3 INTACT-purified endosperm of Col × Ler reciprocal crosses. (XLSX 904 kb) profiling of the endosperm implies exclusion of polycomb group protein targeting by DNA methylation. PLoS Genet. 2010;6(10):e1001152. 8. Moreno-Romero J, Jiang H, Santos-Gonzalez J, Kohler C. Parental epigenetic Acknowledgements asymmetry of PRC2-mediated histone modifications in the Arabidopsis Sequencing was performed by the SNP&SEQ Technology Platform, Science endosperm. EMBO J. 2016;35:1298–311. for Life Laboratory at Uppsala University, a national infrastructure supported 9. Simon JA, Kingston RE. Occupying chromatin: polycomb mechanisms for by the Swedish Research Council (VRRFI) and the Knut and Alice Wallenberg getting to genomic targets, stopping transcriptional traffic, and staying put. Foundation. Mol Cell. 2013;49(5):808–24. Moreno-Romero et al. Genome Biology (2019) 20:41 Page 11 of 11 10. Mozgova I, Hennig L. The polycomb group protein regulatory network. 33. Zhou S, Liu X, Zhou C, Zhou Q, Zhao Y, Li G, Zhou DX. Cooperation Annu Rev Plant Biol. 2015;66:269–96. between the H3K27me3 chromatin mark and non-CG methylation in 11. Luo M, Bilodeau P, Dennis ES, Peacock WJ, Chaudhury A. Expression and epigenetic regulation. Plant Physiol. 2016;172(2):1131–41. parent-of-origin effects for FIS2, MEA, and FIE in the endosperm and 34. Wolff P, Jiang H, Wang G, Santos-Gonzalez J, Köhler C. Paternally expressed embryo of developing Arabidopsis seeds. Proc Natl Acad Sci U S A. 2000; imprinted genes establish postzygotic hybridization barriers in Arabidopsis 97(19):10637–42. thaliana. Elife. 2015;4. https://doi.org/10.7554/eLife.1. 12. Hsieh TF, Shin J, Uzawa R, Silva P, Cohen S, Bauer MJ, Hashimoto M, Kirkbride 35. Bilodeau S, Kagey MH, Frampton GM, Rahl PB, Young RA. SetDB1 RC, Harada JJ, Zilberman D, et al. Regulation of imprinted gene expression in contributes to repression of genes encoding developmental regulators and Arabidopsis endosperm. Proc Natl Acad Sci U S A. 2011;108(5):1755–62. maintenance of ES cell state. Genes Dev. 2009;23(21):2484–9. 13. Law JA, Jacobsen SE. Establishing, maintaining and modifying DNA 36. Yuan P, Han J, Guo G, Orlov YL, Huss M, Loh YH, Yaw LP, Robson P, Lim B, methylation patterns in plants and animals. Nat Rev Genet. 2010;11(3):204–20. Ng HH, et al. Eset partners with Oct4 to restrict extraembryonic trophoblast 14. Pignatta D, Erdmann RM, Scheer E, Picard CL, Bell GW, Gehring M. Natural lineage potential in embryonic stem cells. Genes Dev. 2009;23(21):2507–20. epigenetic polymorphisms lead to intraspecific variation in Arabidopsis 37. Mozzetta C, Pontis J, Fritsch L, Robin P, Portoso M, Proux C, Margueron R, gene imprinting. Elife. 2014;3:e03198. Ait-Si-Ali S. The histone H3 lysine 9 methyltransferases G9a and GLP 15. Shu H, Nakamura M, Siretskiy A, Borghi L, Moraes I, Wildhaber T, Gruissem regulate polycomb repressive complex 2-mediated gene silencing. Mol Cell. W, Hennig L. Arabidopsis replacement histone variant H3.3 occupies 2014;53(2):277–89. promoters of regulated genes. Genome Biol. 2014;15(4):R62. 38. Fukuda A, Tomikawa J, Miura T, Hata K, Nakabayashi K, Eggan K, Akutsu H, 16. Roudier F, Ahmed I, Berard C, Sarazin A, Mary-Huard T, Cortijo S, Bouyer D, Umezawa A. The role of maternal-specific H3K9me3 modification in Caillieux E, Duvernois-Berthet E, Al-Shikhley L, et al. Integrative epigenomic establishing imprinted X-chromosome inactivation and embryogenesis in mapping defines four main chromatin states in Arabidopsis. EMBO J. 2011; mice. Nat Commun. 2014;5:5464. 30(10):1928–38. 39. Inoue A, Jiang L, Lu F, Zhang Y. Genomic imprinting of Xist by maternal 17. Calarco JP, Borges F, Donoghue MT, Van Ex F, Jullien PE, Lopes T, H3K27me3. Genes Dev. 2017;31(19):1927–32. Gardner R, Berger F, Feijo JA, Becker JD, et al. Reprogramming of DNA 40. Inoue A, Jiang L, Lu F, Suzuki T, Zhang Y. Maternal H3K27me3 controls DNA methylation in pollen guides epigenetic inheritance via small RNA. Cell. methylation-independent imprinting. Nature. 2017;547(7664):419–24. 2012;151(1):194–205. 41. Zhang X, Yazaki J, Sundaresan A, Cokus S, Chan SW, Chen H, Henderson IR, 18. Choi Y, Gehring M, Johnson L, Hannon M, Harada JJ, Goldberg RB, Shinn P, Pellegrini M, Jacobsen SE, et al. Genome-wide high-resolution Jacobsen SE, Fischer RL. DEMETER, a DNA glycosylase domain protein, mapping and functional analysis of DNA methylation in Arabidopsis. Cell. is required for endosperm gene imprinting and seed viability in 2006;126(6):1189–201. Arabidopsis. Cell. 2002;110:33–42. 42. Cheadle C, Vawter MP, Freed WJ, Becker KG. Analysis of microarray data 19. Jullien PE, Kinoshita T, Ohad N, Berger F. Maintenance of DNA methylation using Z score transformation. J Mol Diagn. 2003;5(2):73–81. during the Arabidopsis life cycle is essential for parental imprinting. Plant 43. Moreno-Romero J, Santos-Gonzalez J, Hennig L, Köhler C. Applying the Cell. 2006;18(6):1360–72. INTACT method to purify endosperm nuclei and to generate parental- 20. Ebbs ML, Bender J. Locus-specific control of DNA methylation by the specific epigenome profiles. Nat Protoc. 2017;12(2):238–54. Arabidopsis SUVH5 histone methyltransferase. Plant Cell. 2006;18(5):1166–76. 44. Del Toro-De Leon G, Köhler C. Endosperm-specific transcriptome analysis by 21. Wuest SE, Vijverberg K, Schmidt A, Weiss M, Gheyselinck J, Lohr M, Wellmer applying the INTACT system. Plant Reprod. 2018. https://doi.org/10.1007/ F, Rahnenfuhrer J, von Mering C, Grossniklaus U, et al. Arabidopsis female s00497-018-00356-3. gametophyte gene expression map reveals similarities between plant and 45. Clough SJ, Bent AF. Floral dip: a simplified method for agrobacterium- animal gametes. Curr Biol. 2010;20(6):506–12. mediated transformation of Arabidopsis thaliana. Plant J. 1998;16(6):735–43. 22. Belmonte MF, Kirkbride RC, Stone SL, Pelletier JM, Bui AQ, Yeung EC, 46. Moreno-Romero J, Del Toro-De León G, Yadav VK, Santos-González J, Köhler Hashimoto M, Fei J, Harada CM, Munoz MD, et al. Comprehensive C. Epigenetic signatures associated with imprinted paternally-expressed developmental profiles of gene activity in regions and subregions of the genes in the Arabidopsis endosperm. Datasets. Gene Expression Omnibus. Arabidopsis seed. Proc Natl Acad Sci U S A. 2013;110(5):435–44. 2019. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE119915. 23. Stroud H, Greenberg MV, Feng S, Bernatavichute YV, Jacobsen SE. Accessed 7 Feb 2019. Comprehensive analysis of silencing mutants reveals complex regulation of the Arabidopsis methylome. Cell. 2013;152(1–2):352–64. 24. Schon MA, Nodine MD. Widespread contamination of Arabidopsis embryo and endosperm transcriptome data sets. Plant Cell. 2017;29(4):608–17. 25. Waters AJ, Bilinski P, Eichten SR, Vaughn MW, Ross-Ibarra J, Gehring M, Springer NM. Comprehensive analysis of imprinted genes in maize reveals allelic variation for imprinting and limited conservation with other species. Proc Natl Acad Sci U S A. 2013;110(48):19639–44. 26. Wolff P, Weinhofer I, Seguin J, Roszak P, Beisel C, Donoghue MT, Spillane C, Nordborg M, Rehmsmeier M, Köhler C, et al. High-resolution analysis of parent-of-origin allelic expression in the Arabidopsis endosperm. PLoS Genet. 2011;7(6):e1002126. 27. Gehring M, Missirian V, Henikoff S. Genomic analysis of parent-of-origin allelic expression in Arabidopsis thaliana seeds. PLoS One. 2011;6(8):e23687. 28. Klosinska M, Picard CL, Gehring M. Conserved imprinting associated with unique epigenetic signatures in the Arabidopsis genus. Nat Plants. 2016;2:16145. 29. Zhang M, Xie S, Dong X, Zhao X, Zeng B, Chen J, Li H, Yang W, Zhao H, Wang G, et al. Genome-wide high resolution parental-specific DNA and histone methylation maps uncover patterns of imprinting regulation in maize. Genome Res. 2014;24(1):167–76. 30. Mozgova I, Kohler C, Hennig L. Keeping the gate closed: functions of the polycomb repressive complex PRC2 in development. Plant J. 2015;83(1):121–32. 31. Du J, Johnson LM, Jacobsen SE, Patel DJ. DNA methylation pathways and their crosstalk with histone methylation. Nat Rev Mol Cell Biol. 2015;16(9):519–32. 32. Alabert C, Groth A. Chromatin replication and epigenome maintenance. Nat Rev Mol Cell Biol. 2012;13(3):153–67.